YouTube has a big climate misinformation problem it can’t solve

Shutterstock

In a polarized world, social media platforms can’t remain neutral.

The US nonprofit Avaaz has a new report out detailing how YouTube is actively spreading climate misinformation to millions of viewers through its recommendation algorithms, including videos with exciting titles like, ahem, “CIA Whistleblower Speaks Out About Climate Engineering Vaccination Dangers and 911.”

The report contains a number of recommendations for the platform, including working with independent fact checkers to identify such videos and remove or demonetize them.

The recommendations make perfect sense — as long as climate misinformation can be reliably and fairly identified in a way that won’t get YouTube embroiled in political or ideological controversies.

But I don’t think it can.

Avaaz

YouTube and other social media platforms do want to limit the spread of misinformation and hate speech, if only to relieve social pressure and defend their reputations. But they want to do so while remaining ideologically neutral, refraining from anything that might appear to be choosing sides in America’s culture war.

Unfortunately, in a polarized and divided US, where even the most basic facts and values are contested, that posture is becoming impossible. The choice between neutrality and fundamental values like respect for evidence and non-discrimination is becoming unavoidable, and private companies like social media platforms, when pressed, will always choose neutrality, for business reasons.

Social media platforms may well be, as Dawn Stover argues in the Bulletin of the Atomic Scientists, bad for the climate. But in this as in other matters, they reflect an epistemic crisis they cannot themselves hope to solve.

I’ve written about America’s epistemic crisis, our growing inability to learn and know things in common, before (see here, here, and here). This time I want to approach it from a slightly different angle, which I hope will help illuminate YouTube’s unsolvable dilemma. It’s a bit of a journey, involving an analogy drawing on the work of psychologist Daniel Kahneman, particularly his seminal 2011 book Thinking, Fast and Slow.

(Kahneman’s work, often alongside psychologist Amos Tversky, won him a Nobel Prize in economics in 2002 and eventually a Presidential Medal of Freedom. Kahneman and Tversky’s work and friendship was the subject of journalist Michael Lewis’s book The Undoing Project. None of them should be held responsible for my analogy.)

We’ll look at the kind of reflective self-regulation that’s important at both the individual and social level, why the institutions meant to do it in the US have become so weak, and why YouTube can’t hope to fill the breach. Stay with me; it will all come full circle.

Fast and slow thinking

The broad thrust of Kahneman’s work was to challenge “homo economicus,” the conventional economic view of human beings as rational interest maximizers. There’s a lot to it, but at its heart is a distinction between System 1 (S1) thinking and System 2 (S2) thinking.

Kahneman described S1 as “fast, automatic, frequent, emotional, stereotypic, and unconscious,” describing what are more colloquially known as “gut reactions.” S2 is “slow, effortful, infrequent, logical, calculating, and conscious,” which is closer to what we tend to conceive of as thinking — taking a step back, slowing down, consciously assessing and reasoning.

S1 reflects your instincts and predispositions, your fears and blind spots, the stuff that is effectively programmed into your neural and endocrine systems by your DNA, your pre- and early postnatal care, and your formative childhood experiences. S1 is where we find what these days are called “implicit biases,” the kinds of preconscious predilections that shape how you react to a situation before “you,” the conscious, thinking you, is even fully aware of what’s going on.

Win McNamee/Getty Images

S2 enables free will and the pursuit of long-term goals

Insofar as we have any ability to regulate or evolve the S1 systems we inherit — insofar as we have any true free will, in the sense of being able to choose our characters and our fates — it is through S2 thinking, our (limited) ability to step back and consider a situation with some degree of objectivity, to assess our own reactions as a third-party observer might, and to take deliberate steps to shape our conduct toward our aspirations.

It is our S1 systems that push us to road rage or binge eat or endless procrastination. All those things feel right in the moment, because they respond to preprogrammed neural and hormonal signals that we had no hand in consciously choosing.

It is only through S2, through perspective-taking and regulating our short-term impulses, that we get work done, develop long-term skills, or stay physically healthy.

Of course, using S2 to successfully regulate S1 isn’t easy; if it were, therapists would go out of business. S1 is a lot more powerful and a lot older, evolutionarily speaking. Most of the time, for most people, S2 serves S1, not the other way around. (As philosopher David Hume put it, “reason is a slave to the passions.”) Very rarely do people question the assumptions and biases they have inherited. More often, S2 is put to use constructing narratives to justify that inheritance, in service of identity formation and reinforcement, not Truth.

It is difficult to use S2 to impose discipline, to hold ourselves to higher, consciously chosen standards rather than reflexively accepting or justifying our gut responses. But that is the work of being a responsible moral and epistemological agent. It’s what grown-ups do.

The reason I’ve gone through all this is that I think there is something analogous to S1 and S2 thinking going on at the collective, national level.

Fast and slow thinking, socially

Socially speaking, the equivalent of S1 thinking (“fast, automatic, frequent, emotional, stereotypic, and unconscious”) is a default to our immediate family/clan/tribal/identity commitments, the places and people and cultural narratives with which we are familiar and instinctively comfortable.

The equivalent of S2 thinking (“slow, effortful, infrequent, logical, calculating, and conscious”) is our ability to take a step back, bracket or restrain our immediate group interests, identify mutual interests across factional lines, and strike non-zero-sum agreements between groups based on a shared set of rules — transpartisan rules, meant to serve not one faction or the other, but to create a pluralist framework in which multiple factions can thrive. (Think, for instance, of international trade.)

Shutterstock

Our (limited) ability to self-regulate and restrain our S1 thinking, our group-level gut instincts, is more or less what has enabled complex global civilization. But always and everywhere, factional thinking, our tendency to pull our circles of trust and care inward, to fear or dominate the Other, threatens S2 goals.

There are occasionally individual leaders with the wisdom to consistently apply S2 thinking, but as pretty much every monarchy and hereditary system in history has shown, if a particular group is in charge long enough, they will inevitably come to rule for their own benefit, while telling themselves they are doing otherwise. Groups are no better than individuals at self-regulating over time.

So, if we can’t trust any particular set of individuals to maintain it consistently, how can S2 thinking be built into society in an enduring way? The answer, such as it is, lies in institutions. We build S2 thinking into the infrastructure of public life to the extent that we enshrine it in institutions and rules that command trust across factional lines.

Take, for example, the institution of professional science. We want to know the truth about the natural world, but we know that individual humans, even highly educated humans, are prone to all kinds of cognitive blind spots, biases, and errors. The best way we’ve found to correct for those cognitive flaws is to create a kind of guild with special rules and procedures meant to counteract them — strict rules of experimentation and analysis, peer review, replication of results, and the like. The idea is that the guild, collectively, will engage in the kind of sustained self-scrutiny and self-correction that individuals often can’t manage on their own, and thus grope its way toward truth.

Science enforces its norms and procedures not only with explicit professional rules, but also with social pressure. Being a scientist means something to scientists; they take it seriously, so other scientists take it seriously too.

We know all too well that science and the sheen of objectivity it carries can be put to horrific uses (see: eugenics). Plenty of recent evidence confirms that science remains a flawed institution. Peer review contains all sorts of implicit biases; replication is spotty at best. Even with its rules and procedures, a lot of nonsense and shoddy work gets through the gates. Science is composed of humans prone to self-serving reasoning just like everyone else. But overall, as a bit of S2 infrastructure, science has served humanity extraordinarily well.

Take another example: journalism. Post-war US journalism sought to become something slightly more akin to science. It formed a kind of loose guild and developed some guild rules meant to serve as collective self-regulation. Reporters had to get confirmation of every fact from two sources; they had to contact anyone mentioned in a story for comment; they had to run stories past fact checkers and editors; they had to publish corrections. These were the workaday rules passed along in local newsrooms to young reporters.

Getty Images/EyeEm

All those rules were designed to counteract the natural cognitive flaws of individual humans, who are prone to subconsciously (and sometimes consciously) shade and shape the stories they tell. The hope was that journalism would provide something on current events closer to objective truth, like science, and enable similar social advances to those science had unlocked.

To say the least, journalism too has always been a flawed institution, its failings even more copious and better known than those of science. It has proven far less resilient than science too, far easier to corrupt. Nonetheless, it was and remains a noble experiment in designing S2 thinking, mechanisms of self-analysis and self-correction, into the social infrastructure.

So it is with the academy, politics, K-12 education, and all our other public institutions. Their role is collective self-regulation, some imposition of S2 thinking on groups naturally inclined toward a variety of S1 -isms. And as flawed as they are, as often as they fail, as badly as they are often bent to the service of culturally dominant groups, humans haven’t found any social S2 mechanism that work better. Only better institutions.

The US was founded on S2 institutions

The US — in theory, if lamentably not in practice — is a nation founded on an S2 system, on explicitly transpartisan rules meant to bind everyone equally. The checks and balances among branches of government are meant to ensure that no one group takes power and rules on its own behalf. Government administrative and legislative procedures are reified habits, ensuring that certain mechanisms of self-correction are automatically in place. The Bill of Rights is just a set of S2 rules, a way of taking certain classes of decisions out of S1 hands.

And just as with science, all this S2 infrastructure is held in place in part through explicit rules, but also in part (much larger part than most people thought, as the Trump years are demonstrating) through norms, unwritten rules enforced via social approbation or disapprobation.

The institutions work on a foundation of social trust. Groups put aside their short-term interests insofar as they believe that they are part of a larger Us, a larger identity, working collectively toward long-term goals that are to everyone’s mutual benefit. Groups and factions enforce the norms and respect the rules if they trust the S2 infrastructure and believe it is working in their interests.

Many people in the US no longer do.

Javier Zarracina for Vox

S2 systems are breaking down in the US

The last 60 years or so of movement conservatism in the US can be seen as a sustained attack on America’s S2 infrastructure.

The story of its attack on journalism — which I discussed in more detail here — is illustrative. In the post-WWII period of liberal ascendency in the US, conservatives became convinced that journalism was dominated by liberals slanting the news in their favor. (Did you know Richard Nixon consulted with none other than Roger Ailes about creating a conservative news channel?)

To put it in terms of my analogy, conservatives claimed that journalism was no longer an S2 institution, a set of transpartisan rules and procedures meant to produce fair, unbiased, and reliable information, but an S1 front, an institution meant to serve one faction’s interests under the guise of transpartisanship.

It can be debated — and has been, exhaustively — just how much validity there was and is to these conservative complaints about mainstream media. Regardless, what became clear is that the complaints were not animated by a desire to strengthen journalism as an S2 institution. Conservatives did not want less bias or more rigorous fealty to transpartisan rules. When their time came to create a news network of their own, they created Fox, a funhouse-mirror version of journalism with few self-correction mechanisms (fewer every year) and a mission to strengthen the exclusionary S1 identity of a particular demographic.

Fox was, in comically exaggerated form, exactly what conservatives had accused the mainstream media of being: a propaganda outlet pretending to be “fair and balanced,” an S1 faction posing as an S2 institution. The media world has since become split between a mainstream media that goes on believing it is an S2 institution, unbiased and transpartisan, and a growing right-wing media universe convinced that the MSM is liberal and that there are no S2 institutions, only competing factions, their truth or ours.

Thus have conservatives assaulted all of the S2 institutions in American life. In a 2010 quote I never tire of using, conservative radio pioneer Rush Limbaugh called government, academia, science, and media the “Four Corners of Deceit.” All of those institutions are S1 posing as S2, he said, entirely taken over by liberals, and “everything run, dominated, and controlled by the left here and around the world is a lie.”

RushLimbaugh.com

I have called the resulting closed loop “tribal epistemology,” the elevating of group interests over transpartisan principles of evidence, reasoning, and verification, so that what’s good for Us becomes what’s is and must be true.

As a result of this long assault (and many other factors, including their own failings), trust in US institutions, and social trust generally, is at a historic low. Something like 30 to 40 percent of the country has been absorbed into a right-wing bubble of counter-institutions meant to produce, in Kellyanne Conway’s memorable phrase, “alternative facts.” There are things within the conservative movement that look like journalism outlets and think tanks, but they operate on a purely S1 basis, to serve factional interests. The federal government arguably fits that description these days.

The question is, if only one of America’s two warring factions believes the information that mainstream institutions produce, abides by their rules, or socially propagates their norms, are they still S2 institutions at all? Are there still S2 institutions?

We know what an individual with weak or unregulated S2 systems looks like: a drifting set of impulses, compulsions, fears, and hungers, forever reacting, with no ability to step back, get perspective, and self-regulate in service of long-term goals. (See: Donald Trump.) Most people like that don’t live happy lives. Or become president.

What becomes of a nation that loses all capacity for S2 self-regulation, when its constituent groups are no longer able to abide by a shared set of rules, pursue a shared set of long-term goals, or even accept a shared set of facts?

We’re finding out. And that brings us, at last, back to Avaaz and YouTube.

Social media platforms are being put in an impossible position

Social media platforms have been under attack for allowing the spread of misinformation and hate speech, from Facebook’s alleged role in Myanmar’s Rohingya genocide to YouTube’s spread of alt-right videos to Twitter’s hosting of frequent harassment campaigns. Critics, including Avaaz, are pressuring the platforms to do more collective S2 thinking, to rule some content misleading or out-of-bounds.

It is putting the platforms in an impossible situation, stuck between two incommensurate goals. As journalist Neima Jahromi puts it in her New Yorker story on YouTube, CEO Susan Wojcicki wants to “avoid accusations of ideological bias while also affirming her company’s values.”

Asa Mathat for Vox Media

On one hand, as Google CEO Sundar Pichai told Congress in 2018, “It’s really important to me that we approach our work in an unbiased way.” Social media platforms don’t want to be seen as taking sides in political or ideological disputes — in particular, they don’t want to run afoul of Republicans, who run the federal government at the moment.

That means, as Wojcicki recently wrote in a blog post, the platform leans toward openness, even if that sometimes means tolerating “content that is outside the mainstream, controversial, or even offensive.” It is in YouTube’s political and economic interest not to alienate or exclude anyone in particular; everyone is potential ad revenue.

On the other hand, YouTube does affirm values beyond openness. It routinely removes videos that violate its policies and has sought to cut back on “borderline content and harmful misinformation.” Last year it implemented a range of new policies intended to reduce misinformation, including boosting “trusted” news sources and putting pop-up on cards with Wikipedia content on videos about certain subjects like the moon landing and the Holocaust.

Earlier this year, it removed thousands of videos, specifically videos “alleging that a group is superior in order to justify discrimination, segregation or exclusion based on qualities like age, gender, race, caste, religion, sexual orientation or veteran status.” It has recently taken anti-vaccination videos off its ad lists, such that they can’t be monetized.

So YouTube wants to remain ideologically neutral, but also affirm basic epistemic and moral values. Unfortunately, in a country high on polarization and low on social trust, with tattered and weak S2 institutions, that is no longer possible.

Take the climate misinformation videos Avaaz identified.

YouTube

Who decides what is true on climate change?

Some of the videos dispute that CO2 has any role in climate change. Avaaz calls that misinformation. On what basis? Well, Avaaz trusts the scientific community, which overwhelmingly agrees that CO2 drives climate change. In Avaaz’s worldview and value set, science is an S2 institution and climate science, as part of it, warrants trust.

But the people who are disputing CO2’s role in climate change see climate science as an S1 front, not to be trusted. They believe that the global climate scientific community is dominated by liberals, who give each other grants and research money based on fealty to predetermined conclusions. They believe that establishment climate science is just dressed-up liberalism.

And they are backed up in this belief by one of the nation’s largest and most active political factions and one of its two political parties. Republican Sen. Ted Cruz of Texas was on MSNBC host Chris Hayes’s podcast recently reciting climate denialist myths that date back to the 1980s, including some that question CO2’s role.

The conclusion is unavoidable: to trust establishment climate science as the arbiter of truth on climate change is to take a side against conservatives.

YouTube will have to make that choice again and again. It can turn to “independent fact checkers,” but that just moves the problem back a level — then the fact checkers have to make the choice.

Each time a person, whoever it is, is confronted with that choice, they can take one of two paths.

The first is to trust climate science, thereby defining these videos as misinformation worthy of being blocked or demonetized. Doing that will mean taking sides, which will inevitably draw accusations of bias and threats of reprisal.

Sure enough: “In April, Ted Cruz held a Senate subcommittee hearing called ‘Stifling Free Speech: Technological Censorship and the Public Discourse’,” Jahromi reports. “In his remarks, he threatened the [social media] platforms with regulation; he also brought in witnesses who accused them of liberal bias.”

Alex Wroblewski/Getty Images

Conservatives have run this con on one institution after another, battering it with complaints of liberal bias until it is cowed and overreacting in the other direction. It worked on mainstream journalism and now it looks like social media platforms are falling for it as well. YouTube lists Fox News as a “trusted source.” Facebook lists Check Your Fact, a subsidiary of the right-wing Daily Caller, as a “trusted fact checker.” Right-wing misinformation can not be excluded when right-wingers comprise a third of the population and run the federal government.

The second path is to bow to this kind of pressure and accept that carbon dioxide’s role in climate change — and, by proxy, the trustworthiness of establishment climate science — is a subject of legitimate controversy, and that both sides deserve to be heard. Then the misinformation spreads.

This is the basic structure of the problem: if YouTube leaves the videos up, it allows some portion of its audience to be deceived; if it takes them down or demonetizes them, it steps in the middle of an ideological battle against an activist movement, a political party, and at least at the moment, the president himself. There’s no third choice, no way both of Wojcicki’s objectives can be met at once.

And that’s just the clear-cut cases. Beyond overt science denialism, what about videos arguing that climate change damage projections are exaggerated, or that the climate is less sensitive to emissions than conventionally assumed, or that climate change is not as bad a problem as poverty or hunger? Where is the line between misinformation and legitimate dissent? Do we want YouTube drawing it?

YouTube is ill-equipped to play an S2 role

These same kinds of questions have arisen, and will arise, around dozens of other problems. What counts as hate speech? Online shock jock Steven Crowder directed months of racist, homophobic abuse at my colleague Carlos Maza. First YouTube did nothing, then it announced it wouldn’t remove Crowder’s videos, then it announced that it would prevent them from being monetized. (No one ended up happy.)

People who believe that racism and homophobia are bad felt that YouTube wasn’t taking its commitment to inclusion seriously; conservatives felt that YouTube was biased against them for what they claimed were “just jokes.”

There will be dozens, hundreds of similar factual and moral questions facing YouTube (and Facebook, Twitter, etc.) in coming years — what is misinformation, what is hate, where is the line. These are important decisions defining the boundaries of US culture.

It seems clear that social media platforms can’t avoid making them, but equally clear that the chances of them doing the job well are vanishingly slim.

For one thing, YouTube “now attracts a monthly audience of two billion people and employs thousands of moderators,” Jahromi reports. “Every minute, its users upload five hundred hours of new video.” That may quite simply be too much to moderate; it may be that the business model society has allowed social media companies to establish precludes any serious, thoughtful filtering.

But lack of effective moderation doesn’t seem good either. There’s a certain school of thought, dating to Enlightenment ideals, that says more speech is always better, that the answer to bad speech is good speech. But that view appears regrettably naive in the age of information overload.

It has become all too clear that, if anything, misinformation and hate spread faster than trustworthy information. What best drives “engagement,” the holy grail of social media, are negative, brain-stem reactions: outrage, fury, contempt. If there is no governing principle in the algorithm beyond serving what people are most likely to click on in the moment, then at least some very large fraction of users will sort themselves into bubbles of misinformation or hatred. (See this harrowing New York Times story of a young man getting sucked into alt-right YouTube.)

Stopping the spread of hate and misinformation requires moderation. It requires S2 gatekeepers. But those gatekeepers cannot, by definition, be “neutral.” And therein lies the rub.

As private companies, social media platforms will always be more averse to controversy than they are committed to limiting misinformation. A sufficiently large and organized political faction will always be able to bully them into treating any act or fact as “controversial” and thus not their place to adjudicate.

Of all the institutions in American life, private social media companies are the least likely to hold the line against social dissolution. They cannot solve America’s epistemic crisis, they can only reflect it.

The only hope for the values of liberal democracy is to improve and strengthen S2 institutions

It’s good that Avaaz and other groups are pressuring social media companies, but the fact that citizens of a democracy are petitioning multi-billion dollar corporations to restore respect for truth and decency is not a good sign.

Just as an individual can only pursue long-term goals by using S2 thinking to shape, direct, and correct the flaws of S1 thinking, so too a pluralist democracy can only prosper in the long-term by designing some kind of S2 infrastructure — institutions, laws, rules, procedures, and norms — to prevent its various internal factions and identities from falling into zero-sum struggle.

A country without functioning S2 infrastructure, without a frontal cortex, ceases to be a nation. It becomes an erratic collection of factions locked in zero-sum struggle, reacting to situations as they arise, unable to look ahead or effectively plan — unable, for instance, to implement a coherent multi-decade policy program to both prepare for and prevent the worst of climate change.

In a situation of all-against-all struggle, groups without access to economic and social power pay the price. The little people, the outsiders, the vulnerable demographics suffer the most. Ultimately, the Founding Fathers’ central insight, however poorly it has been implemented, remains valid: Pluralism and freedom can only be maintained in the long run by “rule of law, not of men,” rule of S2 thinking, not of factional S1 impulses.

It is transpartisan institutions and transparent rules, equitably enforced, that create social trust in a pluralist democracy. And it is social trust that enables such S2 infrastructure to persist. When it’s working, it creates a positive feedback loop.

But right now it isn’t working. The opposite loop is running, thanks to the many failures of US institutions and the coordinated right-wing backlash against them. Social trust is declining; as it does, institutions have a harder time operating effectively; as they stumble, it further reduces social trust.

If anything can check that negative feedback loop, it’s not going to be social media companies writing better algorithms. Those who believe in liberal values, who respect science and expertise and seek a peaceful, equitable multi-ethnic democracy, must devote themselves to taking power and reshaping public institutions to be more effective, trustworthy, and democratic.

Building S2 infrastructure that lives up to the lofty ideals in America’s founding documents is long, slow work. But it is the only path back from the abyss.